arc42: Von statischer zu dynamischer Dokumentation & arc42 for AI Systems

Talk 1: arc42: Von statischer zu dynamischer Dokumentation

Architektur ist mehr als nur Seiten in Dokumenten – sie ist ein Diagramm aus Entscheidungen, Strukturen und Einschränkungen. Wir zeigen Ihnen, wie Sie mit Confluence, docToolchain und Notion eine vernetzte arc42-Dokumentation implementieren können. Mit Metadaten gesteuerten Tabellen, Querverweisen und wiederverwendbaren Vorlagen zeigen wir Ihnen, wie Sie statische Dokumente in eine pflegbare Architektur-Wissensdatenbank verwandeln können.

Dieter Baier

Dieter Baier is an experienced IT consultant at envite with extensive expertise in agile software development and architecture. His core competencies include application development, particularly focusing on system integration based on modern architectures like Clean Architecture, DDD, and the creation of clear, easily accessible architecture documentation.

Luc Weinbrecht

Meet Luc Weinbrecht, Green BPM Consultant at envite. As a distinguished Camunda Champion, he dives into the realm of cutting-edge software architecture within the Camunda framework. With boundless enthusiasm, Luc pioneers novel solutions for business monitoring and takes a serious stance in crafting strategies to significantly reduce the carbon impact of business processes. Join him on the exciting journey towards a greener, more efficient future.

Talk 2: arcAI42: Extending arc42 for AI Systems

arc42 has given us a precise and disciplined way to document software architecture — but AI systems behave fundamentally differently. They don’t stay static. They learn, drift, depend heavily on data quality, and make probabilistic decisions. Yet today, we still document them as if they were ordinary code components. And that’s where things break.
In real projects, we see the same recurring problems: invisible data dependencies, unclear assumptions, models no one can explain, domain language drifting away from training data, unpredictable deployments, missing ownership, and no architectural place to describe feedback loops, retraining logic, or drift mitigation. When these elements aren’t captured architecturally, AI systems become unmaintainable and impossible to govern.
arcAI42 introduces a small set of lightweight extensions to arc42 that make these learning systems architecturally visible. It adds a Data and Domain Context that describes data sources and meaning, a Model Context that explains model assumptions and provides transparency into model behavior, a Lifecycle & Feedback View that shows how a model evolves, and a Governance & Risk section that captures drift, bias, ownership, and mitigation strategies. It also clarifies how models are tested, deployed, monitored, and rolled back.
In essence, arcAI42 transforms AI from a black box into a first-class architectural element — giving architects and data scientists a shared language, and ensuring these systems remain explainable, traceable, and safe as they become part of everyday infrastructure.

Nikita Golovko

Dr. Nikita Golovko is a seasoned Solution Architect with over 16 years of experience in designing scalable, secure, and cost-effective software architectures for industrial and business-critical systems. With a strong academic background and research in machine learning, he bridges the gap between advanced AI technologies and real-world applications on the shop floor. Nikita has led full project lifecycles—from requirements and architecture design to deployment—ensuring compliance with security and industry standards. He is known for translating business needs into robust architectures and aligning cross-functional teams. Passionate about innovation and sustainability, he focuses on applying AI, IoT, and edge computing to drive continuous improvement in industrial environments.